Tissue separation using QuPath

[MGlinka]

Hi everyone,

I am trying to do tissue separation using QuPath - I tried using pixel classification, and although it was successful in identifying tissue such as muscle, it highlighted areas within other regions. Furthermore, if I tried to separate between tissue regions such as muscularis mucosae and muscularis externa, the classifier would highlight all the muscle regardless of the location. Attaching picture below of example tissue of how I would like the tissue to annotate the regions:

I was suggested using object classifier but it requires information from “Cell detection” function which is not useful in this case (at least in my opinion).

Would anyone have any suggestions on how to go about this?

Thanks.

FYI - the staining is of Picro-Sirius Red for collagen.

[Research_Associate (MicroscopyRA)]

Improving the pixel classifier for tumour-stroma classification, impact of spatial features on the classifier

Assuming you have gone through : Pixel classification — QuPath 0.2.3 documentation The gaussian is pretty much your “pixel intensity” feature. Including it along is essentially no spatial information, and only context provided by downsampled pixels (intensity as well). If you want things like local texture, you need the other features as described in the link. In cases like scratch assays (find the area with no cells), I remove gaussian as I do not want the results affected by things like brig…

The pixel classifier is not magic, but if you have enough stain and or structure variation, and enough well varied training data, it can do a lot. That looks doable, but that’s rather easier to say having not tried your particular example.

If it is not working, even with using stain intensity instead of RGB and while using texture features, you may want to look into classifying objects/SLICs, which can have additional measurement types beyond what the pixel classifier can use (Haralick, distance tiles, smoothing, etc).

[MGlinka]

Thanks for the help!

Actually, the way I solved it was by reducing the resolution substantially (to low) and setting the scale to 8 for the classifier to ignore the minor areas.

Thanks again for the suggestions!